Your conjecture of overshooting is quite plausible. That theory does reconcile the discrepant predictions.

I had another look. That does seem to be a factor.

* First, note that P/B values have not been very high for very much of the time since January 2008. So, the sample size is small--any observation you have gets an outsize importance even if its result was a bit random.
* And yes, it seems that, on the few occasions that valuation multiples have been high since then, we have seen overshoots.

For context:
The average ratio of price to peak-to-date known book since January 2008 has been 1.361. Median among those days has been 1.368.
I think market close today was $434.42 per B.
That is 1.639 times currently known book. Definitely higher than has been common in recent years.
The average price-to-peak-known-book one year after any date with price-to-peak-known-book over 1.60 has been 1.182 (!). Technically, that's the situation today. Below average = overshoot in the few instances we've seen. But book value is out of date, and that was only 3.9% of the time.

If we plug in a roughly plausible estimate of Q2 book, say $421,000, we get P/B of 1.548. If the book estimate is good and the price doesn't change in the next few weeks till the statements come out.
Still, we see overshoots in the past on average:
The average price-to-peak-known-book one year after any date with price-to-peak-known-book over 1.53 has been 1.252. 8% of the time.
The average price-to-peak-known-book one year after any date with price-to-peak-known-book over 1.50 has been 1.296. 11.4% of the time.

So: few examples in the past, and in those few situations there was an overshoot, so the method that I used to construct the post-2008 models implicitly concluded that an overshoot was the norm after a valuation level among the top ~10% of samples.

It's more reasonable to assume that P/B will be at least 1.36, and assuming 1.4 wouldn't be a stretch. At the moment even higher numbers seem plausible, but intuitive perceptions of "normal" change with the stock price.

A much simpler prediction method might well be better.
Predict book per share a year ahead, measured in today's dollars. It has grown at around inflation + 7.3%/year for ages. (this millennium).
Predict P/B a year ahead. The average since the crunch has been around 1.37, give or take.
Divide by today's price and you get a one year real total return estimate.
Add your inflation estimate and you get a nominal return estimate, if you want to know the likely nominal price. (important for options, for example)
Not perfect, but a very plausible central estimate unless you have better information. For example, I think Q2 book may come in a little ahead of trend, so I might knock off a percent.

Jim